On Monday afternoon, China tech startup DeepSeek launched its new AI image generator, Janus-Pro-7B. It claims to outperform OpenAI’s DALL-E 3 and Stability AI’s Stable Diffusion.
According to Mashable SEA, this follows the company’s introduction of a more affordable large language model (LLM), providing a strong alternative to OpenAI’s ChatGPT, which shook US-based AI stocks, led by the plummet of Nvidia and Oracle.
According to the company, its Janus-Pro-7B model is an upgrade to DeepSeek’s previous Janus model, launched late last year.
Like DALL-E, Janus-Pro allows users to use text prompts describing a photo or artwork to generate an image in response.
DeepSeek’s technical report said the new model improves Janus by upgrading its training processes, data quality, and model size, leading to more stable images and richer details. Janus-Pro also uses 72 million high-quality synthetic images combined with real-world data to achieve more visually appealing and stable outputs, as reported by Reuters.
Mashable SEA also reported that DeepSeek claims that Janus-Pro can analyse and generate images.
DeepSeek shared examples showing how Janus-Pro improves on the earlier Janus model, stating that “Janus-Pro delivers more stable outputs for short prompts, with improved visual quality, richer details, and the ability to generate simple text.”
/TISG
Featured image by Depositphotos (for illustration purposes only)